Research on Copyright Infringement Issues of Generative Artificial Intelligence

Abstract

This article takes two AI infringement cases involving Ultraman in Hangzhou and Guangzhou as entry points to analyze the infringement disputes of genera-tive artificial intelligence in the stages of data input, model training, and con-tent output. In the Hangzhou case, the platform was found to be liable for con-tributory infringement because users generated infringing content using its tools, and the court recognized fair use during the training stage. In the Guang-zhou case, the platform directly generated infringing content, was determined to be directly infringing, and fair use was not applicable. The differences in the judgments of the two places reflect the liability stratification of "technology tool providers" and "content service providers", as well as the balance struck by the judiciary between technological innovation and copyright protection. It is recommended to add special provisions on fair use of AI training, establish flex-ible judgment criteria such as the "three-step test" and "transformative use", clarify the liability boundaries of different service models, and promote the co-ordinated development of technological innovation and rights protection.